An NLP-based Question Answering Framework for Spatio-Temporal Analysis and Visualization

Abstract

A spatio-temporal analysis system is becoming critical in many disciplinaries to acquire, analyze, and visualize data. However, conducting spatio-temporal analysis often requires certain levels of domain knowledge and experience: on one hand, sophisticated and domain-specific software design makes the analysis difficult for public users; on the other hand, conveying findings from the analysis could result in ineffectiveness and inefficiencies. In this work, we present a Natural Language Processing (NLP)-enabled Question Answering (QA) framework for spatio-temporal analysis and visualization. It allows users to conduct spatio-temporal analysis by speaking or typing questions. Interactive visualization component in the framework creates better communication between insights and users. We use a dataset from the domain of climate science as a case study to demonstrate the framework. The case study is evaluated through a mid-size software company, and great feedback was received. With the microservice architecture in it, the framework is general enough to be applied in a variety of applications and domains.